FEA Wing Structural Analysis 4

[chyriso]

I'm a aeronautical engineer working in Australia.
Lately I'm working on the FEA wing analysis for my company.

As most of you know, finite element method uses
different approach from classical engineering bending
method. In other words, results from FEA and classical
bending theory always contains small amount of
difference depending on how the model has been done.

Becuase no structural test is gonna performed to
validate the FEA model, we run into the difficulty of
proving that we have got a correct/good FEA model.
Comparing FEA to classical bending theory simply
couldn't get differences under 10%, as expected.

does anyone know how to justify a FEA wing model without doing any structural test? If yes, please let me know. Any recommendation is appreciated.

 

[GregLocock]

You have a few options:

Do a modal test.

FE Model a similar existing wing and test it as well.

Build a scale model of the wing and test it.

Confirm your FEA results for some cases by hand calculations the traditional way.

To be honest, asking questions like this shows you might be rather ambitious in hoping to go from FEA to production without testing.

Incidentally my most recent FEA model came out within 8% of the real thing when it was (eventually) tested, and after refinement (not cheating) we got that to 4%.


Cheers

Greg Locock

 

[chyriso]

Hi Greg,

Thanks for your reply.

In fact, we have done a simple structural test on the wing by applying a point load(two tests, vertical test for simulating lift load and horizontal test for simulating drag and torsion loads) on the tip float (270" from the fuselage centerline). However, we concluded that the load we have applied is too small compare to the limit airload that gonna be seen by the wing in service. We couldn't increase the test load because we were already having the test load at 70% of the float limit load. On the other hand, small test load like this doesn't set the wing skin in buckle(if any), which then hard to justify the stiffness of the FEA wing under high up-bending airload where wing skin may be buckled and effectiness of the skin in carrying bending load must be closely modelled

FYI:
1. FEA model with the same test load shows around 10% error for direct stress, but huge error for torsional/shear stress]

2. Doing a structural test for the wing gonna cost too much for a project which only meant to increase the MZFW for three airplanes - i.e. the client didn't want to spend that much on this project.

Hand calculation using classical engineeering theory has been done and the results show about 5-20% different in bending analysis, depending on which wing station are of interest. What percentage different do you normally take as 'close enough' when comparing between hand cal. and FEA ?

cheers,

Yon

 

[GregLocock]

I was happy with 8% and very happy with 4% for deflections, not stresses.

What's your factor of safety? If it is 4 then a 20% error should be OK, if it is 1.1 then obviously it isn't.



Cheers

Greg Locock

 

[GregLocock]

If you have a real physical wing (which you seem to) and you are unable to test it structurally to your satisfaction (I don't understand why) then a non destructive test that can be used to calibrate your FE model is experimental modal analysis.

Your local university should be able to point you in the right direction.



Cheers

Greg Locock

 

[chyriso]

Air loads(lift, drag, pitching moment) are pressure loads acting continuously across the span of the wing. Due to the spanwise and chordwise distributions of the pressure loads, these loads couldn't be easily stimulated using simple jack or point force loader. Making a test rig for this application may cost millions of dollars.

Unlike automotive industry, aeronautical industry uses a very low margin of safety (i.e. allowable/applied - 1). for this project we only have margin around 10% in most structures. That's why accuracy is crucial to us.

cheers,


Yon

 

[crackman]

Yon

I take it from your statement that your are working on a MZFW increase, that this is an aircraft already in service and has therefore already been certified for the base configuration? If so, the first place to start is to find any referenceable data on the basic aircraft. It is tough to find sometimes but there is alot of data out there and you may even find some test data or prior strain surveys.

First off, it is generally very difficult to get aircraft certified without some test correlation. In fact, the "unwritten" FAA rule is that they require correlation of FEA's in support of structural substatiation to not exceed +- 10%. Note that this includes even being conservative. The stance today is that if you are off 10% either way, the load may be going else and therefore your load path assumptions are incorrect. Assuming that your certification agency is even willing to allow you to do this without a specific correlation, here are two suggestions:

1: correlate the method of your FEA idealization to another test and demonstrate your assumptions for load path, stiffness, effective area, etc. Even though the test is not the one for the MZFW, the real intent is to validate the method.

2: correlate your results to a wing unit beam type analysis. You mentioned a bending analysis but I am suggesting something more like Boeing's TES170 program (still used today for sizing). Many OEM's still use these codes (all generally based on the unit beam method or similar to Bruhn's wing section analysis) for wing analysis rather than building a FEM. Note that the use of these methods are generally questioned far less by the certification authorities as they typically have had a much higher level of validation and have been in use for decades. You should be able to get well within 10% correlation, if you are not, then your unit beam method is not refined enough. Boeing's version is fairly elaborate (takes into account buckling effectivity) but very accurate that is why they still use it for wing box sizing.

Hope this helps and good luck

James Burd
FAA DER - Structures/Fatigue and Damage Tolerance

 

[chyriso]

Hi Crackman,

Yes, you're right. The aircraft is already in service and the client basically want to increase the MZFW without adding any extra strutures to the wing - which is quite a challenge/dream provided we already know that the margin we have on the current MZFW is hitting 0.1 or less when calculated using classical theory. In fact, we are hoping that FEA can provide a more accurate MS and therefore push the limit of the wing structure slightly further towards the limit. However as you might know, we run into trouble of justifying the FEA method that we use, 'cos it's not a conventional method and the benefit earns from FEA analysis normally jeopardised by the error (i.e. incapable of showing good correlation).

Australian airworthiness authority, CASA, allows us to do it without doing any structural testing to neither limit or ultimate load, provided we are able to provide 'convincing arguments'.

Questions that I have are:

1. As you might know from the previous thread, we have done a simple test by applying a horizontal/vertical point load at the tip float. Referring to your suggestion 1, are you suggesting that we could use the same methodology, then just justify the methodology rather than the real load? e.g. (1) model wing skin effectiveness area for low stress when doing FEA model for the test case, then change the skin effectiveness area for the limit case and so long as the test FEA model is right the high stress limit case FEA should be right. This is what I thought, but then, the question I was asked is, "At low load level the wing doesn't buckle, but at limit load level the wing does buckle and the load path might then change, how do you justify that?".

2. Could you tell me where can I get a copy of the Boeing's TES170 program ? I tried to search it online but I got no hit. Is it under Boeing's IP ?

thanks for your reply.


Yon

 

[GregLocock]

You don't have to apply a realistic (multi point)load to validate your structure/FEA, if you are operating in the linear range. You could prove that for say simple bending, simple torsion, and a few other cases that your deflectoins were within 10% over a wide grid of points.


This does not deal with stress raisers and so on, deflection based approaches cannot help much there.





Cheers

Greg Locock

 

[SuperStress]

For simple point loads (known magnitude and location) your correlation should be well within 10%. I performed a FEM correlation on a large widebody aircraft undergoing modification. Known point loads on the ground correlated generally within 5%. What you need is a series of point loads that put the wing in different states of bending and torsion to validate your load paths.

One important thing to consider in any correlation effort is to understand the state of stress in your strain gages when they are zeroed. Is the plane sitting on the gear at 1g? Is there unusable fuel in the wing? Are there other objects/systems/equipment that contribute to the mass of the wing which are not accounted for in your model? This is not the same as the zero-g state your FEM assumes. Consider how this will affect your results.

Once your load paths are well established, you can make the jump to flight loads. How well do you know your flight loads? Are they within 10% (both magnitude and distribution)? How are you modeling fuel slosh loads if it is a wet wing?

Another thought, about conservatism. For any given set of external loads, there is only one balanced solution. If predicted loads are higher in one area ("conservative") they are necessarily lower ("unconservative") in another. Your goal is not to have a conservative model.

Another thought, about your linearity assumption. You seem to be focusing on wing skin buckling, but that may not be your first non-linear effect. Many spars are designed to be intermediate diagonal tension beams, with web buckling starting at relatively low loads. Don't neglect this possibility when trying to correlate. Loading to 70% of DLL would probably have buckling in IDT beams.

I would be very wary of any FEM trying to "fine tune" the analysis inside of a 10% margin of safety without a full limit load test as a minimum. What kind of design details are you omitting from your model for the sake of model size? Small cutouts? System penetrations? Are there other potential stress risers that could precipitate a failure?

Having been on both sides of the certification fence, I don't think I could approve your project, as described, as meeting the intent of the FARs (for U.S. certification). I don't know enough about CASA to know what they mean by "convincing arguments".

Good luck with your effort, but I would start discussions immediately with your certification authority about what they are going to require to certify your modification.

SuperStress

 

[crackman]

I totally agree with Superstress. Without naming names, there once was an OEM who was "encouraged" by the certification authorities to perform an ultimate test on their wing which they did not entirely think was necessary. Three to four years of negotiations later, $$$$ later, multitudes of FEM non linear runs later, volumes of comparisons later, etc., the test was finally performed (the engineers wanted to do it, but management just didnt want to pay the money but finally caved in). The wing failed well prior to reaching ultimate load due to a detail which was overlooked and not predicted by FEM! (note this occurred recently, not in ancient history - history tends to repeat itself lest we learn our lessons and behave humbly once in a while)

FEMs are great tools but they must (and I cannot overemphasize this) be based on correlation to test data. This does not mean one must test everytime, however, one must at least run thru the entire exercise once to understand it and to understand the design and load path of the structure they are modelling. In the old days, OEMs tested all the time and engineers benefited and were trained with that knowledge. The only two OEMs in the USA who still feel this way (although probably not much longer now that beancounters run them) is Boeing and Cessna.

Now, obviously, your case is a bit different. You are not designing a new wing, only a modification. This lessens the overall magnitude of the effort but not its importance.

Superstress is entirely right about the correlation. Linear behavior can be easily correlated to, however, its the non-linear behavior that is difficult to predict. Superstress' example of the spar is a very good one. I am currently working an aging aircraft program where the rear spar web buckles at about 75% of limit load.

Considering your currently low margins with the existing MZFW using classical methods, I am not sure FEM will get you there without some sort of correlation to test data representative of actual flight conditions.

Now, that being said, I would not rule out entirely the possibility of refining your classical methods. There can be a large variation in classical results considering which assumptions are made. The Boeing TES170 method was well correlated to testing and provided very good results (I do know it was used on the 707, 727, 747, and for sizing on the 757 and 767 I believe). So, FEMs are not entirely better but do provide ways of handling things other than a simple box beam.

Also, obviously not being the OEM and probably not having access to their data, you might not know this. But, do you know what structure is critical for changes in the MZFW? Most OEMs generally do this by comparing VMT envelops for the wing and in some cases even are able to write some of it off as not being critical without having to reanalyze. But alas you do not have this luxury. So, you may want to revisit how your loads were developed. Depending on how complete your loads analysis is, you may want to look at what load conditions are affected by the increase in MZFW vs current MZFW and which particular structural components are affected. Then,see which areas are designed as shear resistant or not.

Well,this probably has not been much help but I wish you good luck. As for TES170, I do not believe it is available anywhere on the net. The method was documented in old stress reports but not sure it is available. Its essentially a much elaborated on version of whats in Bruhn or Cozzone's paper on Unit Beam (I myself am still looking for an original copy of his paper - 1943 I believe).

Definitely talk to your cert people and lay out a good plan well ahead of time and it will save you in the long run, they are there to help you. The "once upon a time" OEM I mentioned above where the wing failed ended up paying 10 times what the test would have cost had they done it during the desing phase. I am a firm believer that bean counters have their place, but engineers should be running aircraft companies.

Lots of luck

James Burd
FAA DER - Structures/Fatigue and Damage Tolerance

 

[chyriso]

I totally agree that FEM model must be correlated to the test results and structural test must be perform at certain stage of the design as to validate the model - Especially when we already know that the margin of safety is very low and classical theory has proven that increasing the target MZFW will results in stringer buckling failure. Well, this idea has been expressed more thoroughly lately with lots of graphs and change of course of actions has been planned, which is good.

I guess as an engineer sometimes it's hard for us to convince people that they must go through the normal way rather than the shortcut. Facts that might seem obvious to us may be hard to be seen in other's eyes, which is a bummer.